Loss of DNA mismatch repair genes leads to acquisition of antibiotic resistance independent of secondary mutations

TL;DR

Bacteria with broken DNA repair systems become more resistant to antibiotics without needing new mutations, possibly due to increased DNA mixing and cell death.

Contribution

A novel antibiotic resistance mechanism is revealed that arises from loss of DNA mismatch repair genes, independent of secondary mutations.

Findings

Deletion of MMR genes mutL or mutS increases antibiotic resistance in E. coli and Salmonella.

Increased resistance correlates with higher homoeologous recombination rates and cell lysis in MMR mutants.

This resistance pathway allows bacteria to survive antibiotics long enough to develop specific resistance mutations.

Abstract

Antibiotic resistant bacteria have been a major clinical concern for decades. Beyond acquisition of alleles conferring resistance, bacteria under stress (e.g., from changing environmental conditions or mutations) can have higher intrinsic resistance to antibiotics than unstressed cells. This concern is expanded for gram-negative bacteria which have a protective outer membrane that serves as an additional barrier against harmful molecules such as antibiotics. Here, we report a pathway which increases antibiotic resistance (i.e., minimum inhibitory concentration) in response to inactivation of the DNA Mismatch Repair pathway (MMR). This pathway led to increased intrinsic resistance and was independent of secondary mutations. Specifically, deletion of the DNA mismatch repair genes mutL or mutS caused resistance to various antibiotics spanning different classes, molecular sizes, and mechanisms of action in several different E. coli K-12 MG1655 strains, and in Salmonella enterica serovar Typhimurium LT2. This pathway did not change outer membrane permeability or efflux rates. However, the patterns of resistance in MMR mutants correlated with previously reported increases in rates of homoeologous recombination (homologous recombination between non-identical DNA strands). Mutations expected to lower rates of recombination in MMR mutants also decreased the resistance to some antibiotics. Finally, we found lysis occurs in MMR mutants and may contribute to resistance. Our results have demonstrated a novel mechanism that increases antibiotic resistance in direct response to loss of MMR genes, and we propose this resistance involves increased rates of homoeologous recombination and cell lysis. The increased antibiotic resistance of MMR mutants provides a path for these cells to survive in antibiotics long enough to develop more specific resistance mutations and so may contribute to the development of new clinical resistance alleles.

Antibiotic resistance has become a worldwide clinical threat and understanding resistance mechanisms is essential for continued treatment of bacterial infections. Here, we investigate a novel mechanism acting when DNA mismatch repair (MMR) is lost that increases the concentration of many antibiotics needed to kill cells and that does not require secondary mutations. We observed this increase in resistance in both E. coli and Salmonella. Our data suggest increased rates of homoeologous recombination (between non-identical DNA strands) and the lysis of some cells within a population are involved in this resistance. This pathway provides a mechanism for cells with an increased mutation rate due to loss of MMR could survive long enough in the presence of antibiotics to develop new resistance mutations, leading to the spread of antibiotic resistance.

Figures

Increased antibiotic resistance occurs due to loss of MMR genes and not increased mutation rates.(A) Five transductants of ΔmutL::kan and ΔmutS::kan into a ΔyhdP strain were assayed and had comparable levels of vancomycin resistance as demonstrated by efficiency of plating assays (EOPs). (B) Kirby-Bauer tests show increased levels of antibiotic resistance in ΔbamE ΔmutL strains as compared to the ΔbamE strain as demonstrated by smaller zones of clearance. In contrast, black arrows indicate colonies that have acquired spontaneous resistance within the zone of clearance. Scale bars: 2 mm. (C) Quantification of Kirby-Bauer tests demonstrates increased resistance in the ΔmutL ΔbamE strain compared to the ΔbamE strain. Mean of five biological replicates ± the SEM. * p < 0.05 and p < 0.01 by Student’s T test. (D) Loss of MMR genes at different steps in the MMR pathway confers varying degrees of antibiotic resistance. Inhibition of ΔmutL and ΔmutS causes higher levels of antibiotic resistance than wild type. ΔmutH is less resistant in some cases than ΔmutL and ΔmutS, while the degree of resistance with ΔuvrD differs between antibiotics. However, Δdam, which causes the same increase in mutations as loss of MMR, does not confer resistance. Quantification of EOPs normalized to wild type growth is shown as the mean of three biological replicates ± the SEM with individual data points. * p < 0.05 by Mann Whitney test. (E) A Tn10 linked to mutS was transduced into two ΔmutS strains and resulting mutS+ and ΔmutS transductants were tested for their antibiotic resistance. This complementation returned antibiotic resistance to wild-type levels, as assayed by EOP. EOPs are quantitated as in (D).

Increased antibiotic resistance from loss of MMR genes extends beyond E. coli.Deletion of mutL or mutS in Salmonella enterica Serovar Typhimurium LT2 results in increased antibiotic resistance compared to wild type. Thus, loss of MMR genes as a mechanism for antibiotic resistance is not specific to E. coli. EOPs were normalized to wild type and quantitated. Quantification is shown as mean of three biological replicates ± the SEM with individual data points. * p < 0.05 by Mann Whitney test.

Loss of MMR genes does not change outer membrane permeability or efflux rates.(A) ΔbamE, a non-essential lipoprotein in BAM (β barrel assembly machinery), causes decreased resistance to the assayed antibiotics. Combining ΔmutL or ΔmutS with ∆bamE restores partial resistance to these antibiotics as assayed by EOP. EOP for three separate transductants is shown. EOP quantification is normalized to wild type and shown as the mean of three biological replicates ±the SEM with individual data points. (B) PldA is a phospholipase that cleaves phospholipids in the outer leaflet of the OM while, MlaA is the OM component of a system that returns mislocalized phospholipids from the outer leaflet of the OM to the IM. Loss of MMR genes in ΔpldA and ΔmlaA increases antibiotic resistance above that of wild type as assayed by EOP. EOPs were quantitated as in (A). (C) OM permeability was assayed by cleavage of nitrocefin a colorimetric cephalosporin by β-lactamase in the periplasm. The mean of six biological replicates ± the SEM is shown. Loss of MMR genes did not affect the rate of nitrocefin cleavage. (D) To determine whether MMR mutations change rates of efflux a NPN efflux assay was performed. The mean of three biological replicates ± the SEM is shown. Loss of MMR genes did not affect the rate of NPN efflux after glucose addition. (E) To confirm that efflux was not involved in resistance, the effect of deletion of MMR genes on antibiotic resistance was assayed in tolC::Tn10 insertion mutants. EOPs were quantitated as in (A). MMR mutants resulted in increased resistance in the absence of TolC indicating that efflux through the OM is not required. (F) To determine whether heteroresistance occurs in the MMR mutants, E tests were performed with MIC test strips for gentamicin and streptomycin. No colonies in the zone of clearance are observed suggesting that the mechanism observed is not heteroresistance. Images showing MIC markings are shown in S3F Fig. Images are representative of three independent experiments. * p < 0.05 by Mann Whitney test.

Increased antibiotic resistance with loss of MMR genes is linked to homologous recombination.(A) By EOP, resistance from loss of MutL and MutS is greater than that of MutH and often UvrD; however, resistance from loss of both MutH and UvrD is similar to that of MutL and MutS. These changes correlate well with the reported rates of increase in homologous recombination rates MMR mutants. EOP was quantitated and normalized to the wild-type strain. Data are shown as the mean of at least three biological replicates ± the SEM with individual data points. (B) ΔrecA, ΔrecB, and ΔrecC, which reduce homologous recombination rates, result in decreased antibiotic resistance in both the wild type and MMR gene deletion backgrounds for most of the assayed antibiotics. EOPs were quantitated as in (A). * p < 0.05, p < 0.005, * p < 0.0005 compared to wild type; ‡ p < 0.05, ‡‡ p < 0.005 compared to parent strain by Mann Whitney test.

Loss of MMR genes causes increased cell lysis.(A) CPRG is a poorly cell-permeable, colorimetric β-galactosidase substrate that must encounter β-galactosidase, either through increased OM permeability or cell lysis, to be cleaved releasing chlorophenol red. Deletion of MMR genes causes increased CPRG cleavage, suggesting either increased OM permeability or cell lysis. (B) To assay lysis, immunoblots detecting the cytoplasmic protein GroEL were performed on the TCA precipitated supernatant of overnight cultures from an equal culture OD600. Strains with deletions of MMR genes show higher levels of GroEL in the supernatant than wild-type cells do, demonstrating increased lysis. However, this lysis was not decreased by loss of RecB or RecC. Quantification of GroEL levels is shown as mean of three biological replicates ± the SEM. (C) Necrosignaling may play a role in resistance to some assayed antibiotics. AcrA released from lysing cells is sensed by neighboring cells to induce a decrease in antibiotic resistance [82,83]. The change in resistance from loss of necrosignaling (ΔacrA) was compared to that of loss of the AcrAB-TolC efflux pump (ΔacrB). While ΔacrA did not change levels of vancomycin or gentamicin resistance in the ΔmutL and ΔmutS backgrounds, ΔacrA resulted in significant loss of resistance to streptomycin. EOP was quantitated and normalized to the wild-type strain. Data are shown as the mean of at least three biological replicates ± the SEM with individual data points. * p < 0.05, p < 0.005 compared to wild type; ‡ p < 0.05 compared to parent strain by Mann Whitney test.

Full text

Introduction

Antibiotic resistance has become a global threat. Every year millions of infections occur from antibiotic resistant microbes, many of those being infections with gram-negative bacteria [1,2]. In the United States alone, the Centers for Disease Control and Prevention (CDC) estimate there are approximately three million antibiotic resistant infections, resulting in more than 36,000 deaths each year [3]. With the slow development of new therapeutics and microbes constantly adapting to antimicrobial compounds, understanding how bacteria acquire resistance to antibiotics is a pressing issue.

To be defined as resistant to an antibiotic, the minimum inhibitory concentration (MIC) of the antibiotic for the strain must be above a well-defined clinical breakpoint [4]. There are multiple mechanisms microbes that use to become resistant to harmful substrates, including limiting drug uptake, modifying a drug target, inactivating a drug, active drug efflux, and, for some antibiotics, becoming auxotrophic [5]. These mechanisms can be mediated by plasmid acquisition, chromosomal mutations, addition of new genes, or phenotypic adaptation [6]. Furthermore, new resistance alleles can quickly expand in populations through vertical and horizontal gene transfer, becoming widespread. On top of these genetic alterations, the outer membrane (OM) increases the intrinsic resistance of gram-negative bacteria to antibiotics [2,7,8]. Often, mutations that decrease OM permeability can increase resistance to an antibiotic (i.e., the MIC) without crossing the clinical breakpoint to become resistant [7,9–12]. However, these mutations increase the probability of the strain acquiring alleles to become resistant [7,9–12]. In this study, we will describe a phenotype of increased resistance that occurs in DNA mismatch repair (MMR) mutants.

Nucleotide mismatches, often a result of DNA replication, cause alterations to the local structure of the DNA. The MMR pathway corrects nucleotide mismatches as well as other types of DNA damage that cause disruption of DNA structure [13–16]. This pathway requires the protein components MutS, MutL, MutH, DNA helicase II (aka MutU/UvrD), exonucleases, single-stranded binding protein, DNA polymerase III, DNA Ligase, and Dam methylase [17] (S1 Fig). MutS, known as the mismatch recognition protein, recognizes base-to-base mismatches and small nucleotide insertion/deletion mispairings and is ATP-dependent. MutS then recruits MutL to form a complex. After the complex is formed, MutH is recruited and scans 5′ or 3′ to a hemi methylated GATC site. Then, it makes a strand-specific nick of the newly synthesized DNA strand [18]. DNA helicase (UvrD) then unwinds the strand with the nick, followed by degradation of the strand by an exonuclease. Degradation begins at the nick and continues past the mismatch that was recognized by MutS. The single strand will then be resynthesized and ligated by DNA polymerase III and DNA ligase, respectively. Finally, the new strand will be methylated, completing the repair process [15].

Point mutations to MMR genes have been reported in antibiotic resistant clinical isolates of both gram-positive and gram-negative bacteria [19–25]. These strains also had other mutations that were thought to be the cause of the antibiotic-resistant phenotype, with MMR mutations facilitating development of the other mutations due to the lack of strand specificity for mismatch repair by other DNA repair pathways. However, the effect of the MMR mutations on antibiotic resistance in isogenic strains was not evaluated.

In a study characterizing an OM permeability mutant (ΔyhdP), Mitchell, et al. observed that disruption of MMR genes by transposon insertion caused vancomycin resistance in E. coli K-12 [26]. YhdP is a protein which plays a role in phospholipid transport between the inner membrane (IM) and OM [27–29]. Deletion of yhdP causes cells to become much less resistant to vancomycin and SDS EDTA [26]. Transposon mutagenesis was performed in a strain with a deletion of yhdP. Transposon mutants with increased resistance to vancomycin were selected from the transposon library. Transposon insertions in mutL or mutS, inactivating MMR, led to vancomycin resistance in the ΔyhdP strain; however, no effect of these mutations was observed for ΔyhdP’s SDS EDTA sensitivity, suggesting that inactivating MMR affected vancomycin resistance rather than ΔyhdP specific functions [26].

Here, we set out to explore the link between inhibition of MMR and increased vancomycin resistance in E. coli K-12. We have found the resistance extends to many antibiotics beyond vancomycin and is caused directly by loss of MMR without the need for further mutations. It can be observed in wild-type E. coli, in OM permeability mutants, and in Salmonella enterica Serovar Typhimurium. Furthermore, our data suggest the mechanism of this resistance involves increased rates of homoeologous recombination (recombination between non-identical DNA strands) and increased rates of cell lysis. Understanding mechanisms of increased resistance via inhibition of MMR will allow for new discoveries to be made on how to target antibiotic resistance mutants and alleviate the epidemic of antibiotic-resistant infections.

Results

Increased antibiotic resistance occurs directly from loss of MMR genes and not from increased mutation rates

Hypermutators, such as cells with inhibition of MMR function, have been linked to antibiotic resistance [30]. Clinical antibiotic-resistant strains with MMR mutations have been assumed to have increased mutation rates allowing them to gather other favorable mutations, but the direct effects of MMR on antibiotic resistance have not yet been studied. Given that we previously observed increased vancomycin resistance in a more vancomycin sensitive ΔyhdP strain with several independent transposon mutants in MMR genes in E. coli K-12 [26], we first investigated whether the relationship between loss of MMR and increased vancomycin resistance required secondary mutations facilitated by increased mutation rates. If the increased antibiotic resistance were due to secondary mutations, the level of resistance should differ between separately constructed MMR mutants as these strains would have different spontaneous mutations, and these mutations would occur at different times. Therefore, we set out to investigate the effect of deletions of MMR genes on antibiotic resistance in independently constructed strains through efficiency of plating assays (EOPs). We transduced ΔmutL and ΔmutS alleles into a ΔyhdP strain and assayed vancomycin resistance in five separate transductants. We observed the same level of resistance in all transductants (Fig 1A). The equal resistance phenotypes observed in all independent transductants demonstrate that loss of MMR is directly responsible for the resistance observed, and secondary mutations are not necessary for the resistance.

Increased antibiotic resistance occurs due to loss of MMR genes and not increased mutation rates.(A) Five transductants of ΔmutL::kan and ΔmutS::kan into a ΔyhdP strain were assayed and had comparable levels of vancomycin resistance as demonstrated by efficiency of plating assays (EOPs). (B) Kirby-Bauer tests show increased levels of antibiotic resistance in ΔbamE ΔmutL strains as compared to the ΔbamE strain as demonstrated by smaller zones of clearance. In contrast, black arrows indicate colonies that have acquired spontaneous resistance within the zone of clearance. Scale bars: 2 mm. (C) Quantification of Kirby-Bauer tests demonstrates increased resistance in the ΔmutL ΔbamE strain compared to the ΔbamE strain. Mean of five biological replicates ± the SEM. * p < 0.05 and ** p < 0.01 by Student’s T test. (D) Loss of MMR genes at different steps in the MMR pathway confers varying degrees of antibiotic resistance. Inhibition of ΔmutL and ΔmutS causes higher levels of antibiotic resistance than wild type. ΔmutH is less resistant in some cases than ΔmutL and ΔmutS, while the degree of resistance with ΔuvrD differs between antibiotics. However, Δdam, which causes the same increase in mutations as loss of MMR, does not confer resistance. Quantification of EOPs normalized to wild type growth is shown as the mean of three biological replicates ± the SEM with individual data points. * p < 0.05 by Mann Whitney test. (E) A Tn10 linked to mutS was transduced into two ΔmutS strains and resulting mutS+ and ΔmutS transductants were tested for their antibiotic resistance. This complementation returned antibiotic resistance to wild-type levels, as assayed by EOP. EOPs are quantitated as in (D).

Next, we further investigated the resistance phenotype by performing Kirby-Bauer tests (i) to confirm the phenotype we observe is, in fact, resistance (i.e., an increase in MIC), (ii) to determine whether the resistance was vancomycin-specific or broader, and (iii) to investigate whether MMR mutants would confer resistance in OM-deficient mutants beyond ΔyhdP. We measured resistance based on the zone of clearance around the antibiotic disk: the smaller the zone of clearance, the more resistant the strain. Furthermore, this assay allows us to differentiate between changes in resistance of the strain from heteroresistance due to spontaneous resistance mutants or gene amplifications occurring within the population [31] (i.e., colonies within the zone of clearance) (Fig 1B, black arrows). Specifically, we assayed resistance in a strain with a deletion of bamE, a non-essential part of the β-barrel assembly machine (BAM) complex that inserts outer membrane proteins (OMPs) into the OM [32]. This strain shows decreased resistance to several antibiotics that are generally excluded by the OM (**Fig 1BC) [33]. In the ΔbamE background, we observed significantly smaller zones of clearance with ΔmutL for several antibiotics to which bamE mutants have decreased resistance (Fig 1**BC), demonstrating population-level increased resistance to the antibiotics. In addition, these data show that the increased resistance is not restricted to the ΔyhdP background or to vancomycin resistance.

We then asked if resistance occurs in a wild-type background and whether mutants in different steps of the MMR pathway cause equal levels of resistance. Our original screen detected insertions in mutL, mutS and uvrD as vancomycin resistant, but the screen was not saturating [26]. We observed no bulk growth defects in strains with deletions of MMR genes (S2A Fig). We first performed Kirby-Bauer tests with a wide range of antibiotics to determine the antibiotics for which loss of MMR genes might cause a difference in resistance. Then, we performed EOPs and observed resistance to several antibiotics in a wild-type background with ΔmutL and ΔmutS that was very similar for each deletion (Fig 1D). We also confirmed that plasmid-based expression of mutL or mutS could decrease the resistance of their respective deletion strain (S2B Fig). Moreover, transduction of a Tn10 transposon linked to mutS into ΔmutS strains restored wild-type levels of resistance in strains that gained wild-type mutS but not strains maintaining ΔmutS (Figs 1E, S2C). When we assayed resistance to gentamicin and streptomycin by MIC assay in a wild-type background, we also observed a similar increase in MIC for the ΔmutL and ΔmutS strains (S1 Table). These data confirm that loss of MMR genes can cause increased resistance in an otherwise wild-type strain.

In comparison to ΔmutL and ΔmutS, ΔmutH had an intermediate phenotype: mutH was equally or more resistant to antibiotics than the wild type, but less resistant than ΔmutL or ΔmutS strains in some cases (Fig 1D). Deletion of the helicase uvrD caused the strain to be more resistant than the ΔmutH strain, at times similar to the ΔmutL or ΔmutS strains. However, deletion of dam, encoding the Dam methylase responsible for the methylation that directs the strand specificity of MMR, had no effect on antibiotic resistance confirming that increased mutation rates are not sufficient for the resistance. Streptomycin resistance is closely linked to mutations in specific codons of rpsL [34]. To determine whether these mutations occur in MMR strains during streptomycin selection, we performed a Kirby-Bauer test and sequenced the genome of ΔmutL cells taken from the edge of the streptomycin zone of clearance and compared them to unselected cells. We observed no mutations in rpsL (S2 Table), although sequencing suggested 1 bp deletions in the end of bluR, a transcriptional regulator that is not associated with streptomycin resistance [33], and in an intergenic region. To further confirm that resistance mutations were not occurring during selection in MMR mutants, we treated ΔmutS with gentamicin, recovered colonies that survived selection, then used transduction of the Tn10 linked to mutS to restore mutS. We found that the mutS restored cells had wild-type levels of gentamicin resistance (S2D Fig), demonstrating they had not acquired gentamicin resistance mutations during selection.

Thus, although MMR mutations cause a direct increase in antibiotic resistance in a wild-type background, the level of resistance is not equal between the MMR mutants, suggesting that the method of MMR inhibition is important to the resistance. Altogether, we observed direct resistance, not requiring secondary mutations, in a wild-type background and in two different OM permeability mutants. We observed resistance to eight different antibiotics—vancomycin, streptomycin, gentamicin, rifampicin, nalidixic acid, bacitracin, novobiocin and erythromycin. These antibiotics do not share a class, molecular properties (e.g., size, charge), or mechanism of action, demonstrating the breath of this resistance.

Increased antibiotic resistance via loss of MMR extends beyond E. coli.

MMR is an evolutionarily conserved process across many organisms [14]. As we characterized the scope of the resistance caused by loss of MMR genes in E. coli K-12, we turned to whether the resistance extended beyond E. coli or whether it was E. coli specific. Therefore, we investigated whether loss of MMR genes caused increased antibiotic resistance in Salmonella enterica serovar Typhimurium LT2, the type strain for the serovar. Using λ Red recombineering, we replaced the mutL and mutS genes with a kanamycin resistance cassette [35]. We then identified changes in antibiotic resistance by Kirby-Bauer tests and tested the effect of the ΔmutL and ΔmutS mutations in these strains by EOP. The ΔmutL and ΔmutS strains showed increased antibiotic resistance when compared to wild-type S. Typhimurium (Fig 2). The antibiotics for which we observed increased resistance, and the level of resistance observed, were similar to that which we observed with E. coli K-12. These data demonstrate that loss of MMR genes leads to antibiotic resistance beyond E. coli K-12 and suggests that the resistance is due to the same mechanism, since the pattern of resistance is similar between the species.

Increased antibiotic resistance from loss of MMR genes extends beyond E. coli.Deletion of mutL or mutS in Salmonella enterica Serovar Typhimurium LT2 results in increased antibiotic resistance compared to wild type. Thus, loss of MMR genes as a mechanism for antibiotic resistance is not specific to E. coli. EOPs were normalized to wild type and quantitated. Quantification is shown as mean of three biological replicates ± the SEM with individual data points. * p < 0.05 by Mann Whitney test.

MMR’s effect on antibiotic resistance is not due to alteration of the outer membrane permeability barrier or efflux

Changes in antibiotic resistance similar to what we have observed with MMR mutants can be caused by physiological changes to the OM [7,9–12], which serves as an intrinsic barrier against antibiotic entry [12,36,37]. As loss of MMR genes leads to increased antibiotic resistance in a wild-type E. coli strain and a ΔyhdP strain, we next set out to determine whether loss of MMR could increase antibiotic resistance in other OM permeability mutants or whether the resistance was restricted to deficiencies in specific OM pathways, which would suggest a change in OM biogenesis. We identified strains with OM defects causing decreased resistance to antibiotics based on Kirby-Bauer tests. Then, we coupled deletion of the implicated genes with ΔmutL or ΔmutS and performed EOPs for antibiotics to which the OM mutants were less resistant. We began with Bam complex members [32], since the Bam complex has been implicated in OM permeability and serves as an antimicrobial target [38–41]. As mentioned above, BamE is a lipoprotein that serves a bridging function to coordinate activation of BamA and BamD, the essential members of the Bam complex [42]. We observed that the ΔbamE strain showed decreased resistance to vancomycin, rifampicin, and bacitracin (Fig 3A, S3A). When ΔbamE was coupled with either ΔmutL or ΔmutS, we observed increased resistance in the double mutants. This level of resistance was much greater than the ΔbamE parent strain but less than that of wild type. To confirm the resistance was also independent from secondary mutations in the ΔbamE background, we made three individual double mutants and observed similar resistance in each of them (Figs 3A, S3A). We observed similar results in ΔbamB (S3B Fig), which codes for another non-essential BAM lipoprotein [43,44], and bamA101 (S3C Fig), a promoter-down mutant of the gene coding for the essential BamA OMP [45], backgrounds.

Loss of MMR genes does not change outer membrane permeability or efflux rates.(A) ΔbamE, a non-essential lipoprotein in BAM (β barrel assembly machinery), causes decreased resistance to the assayed antibiotics. Combining ΔmutL or ΔmutS with ∆bamE restores partial resistance to these antibiotics as assayed by EOP. EOP for three separate transductants is shown. EOP quantification is normalized to wild type and shown as the mean of three biological replicates ±the SEM with individual data points. (B) PldA is a phospholipase that cleaves phospholipids in the outer leaflet of the OM while, MlaA is the OM component of a system that returns mislocalized phospholipids from the outer leaflet of the OM to the IM. Loss of MMR genes in ΔpldA and ΔmlaA increases antibiotic resistance above that of wild type as assayed by EOP. EOPs were quantitated as in (A). (C) OM permeability was assayed by cleavage of nitrocefin a colorimetric cephalosporin by β-lactamase in the periplasm. The mean of six biological replicates ± the SEM is shown. Loss of MMR genes did not affect the rate of nitrocefin cleavage. (D) To determine whether MMR mutations change rates of efflux a NPN efflux assay was performed. The mean of three biological replicates ± the SEM is shown. Loss of MMR genes did not affect the rate of NPN efflux after glucose addition. (E) To confirm that efflux was not involved in resistance, the effect of deletion of MMR genes on antibiotic resistance was assayed in tolC::Tn10 insertion mutants. EOPs were quantitated as in (A). MMR mutants resulted in increased resistance in the absence of TolC indicating that efflux through the OM is not required. (F) To determine whether heteroresistance occurs in the MMR mutants, E tests were performed with MIC test strips for gentamicin and streptomycin. No colonies in the zone of clearance are observed suggesting that the mechanism observed is not heteroresistance. Images showing MIC markings are shown in S3F Fig. Images are representative of three independent experiments. * p < 0.05 by Mann Whitney test.

To investigate other OM functions, we constructed ΔmlaA and ΔpldA mutants coupled with deletion of MMR genes. PldA is a phospholipase, which contributes to the asymmetry of the OM by cleaving phospholipids that have become mislocalized to the outer leaflet of the OM and initiating a pathway that leads to upregulation of LPS production [46–50]. MlaA is an OM lipoprotein involved in a retrograde phospholipid trafficking pathway that maintains OM lipid asymmetry by removing mislocalized outer leaflet phospholipids and transporting them back to the IM [51–53]. Once again, we observed resistance with either ΔmutL or ΔmutS when compared to the parent OM mutants, although these OM mutants were equally resistant to the wild-type strain (Figs 3B, S3D). Our data demonstrate that the phenotype of resistance seen with loss of MMR genes can overcome multiple kinds of increased OM permeability.

The increased resistance in OM mutants in differing pathways (i.e., phospholipid transport, OMP biogenesis, OM asymmetry) demonstrates the mechanism of increased resistance is not tied to a specific OM biosynthesis pathway, suggesting the resistance may not relate to strengthening of the OM. To confirm this, we tested OM permeability using a nitrocefin cleavage assay. Nitrocefin is a cephalosporin that acts as a colorimetric β-lactamase substrate [54]; thus, it allows permeability through the OM to the periplasm to be assayed [55]. Expression of an open porin mutant of ompC, ompCΔW103-F110 [56], increased OM permeability and so greatly increased nitrocefin cleavage (S3E Fig). In contrast, ΔompR lowering OMP expression [57], showed a lesser increase in cleavage rates suggesting that nitrocefin may also report on other aspects of OM permeability than OMP levels. When we treated wild-type, ΔmutL, and ΔmutS cells carrying a β-lactamase gene with nitrocefin, we observed equal rates of cleavage for each strain (Fig 3C), indicating that OM permeability had not changed.

We further asked whether an increase in efflux rate could be responsible for the resistance we observed. We performed an NPN (N-phenyl-1-naphthylamine) efflux assay [55] where cells were incubated with CCCP (carbonyl cyanide m-chlorophenyl hydrazone) to disrupt the proton motif force and then, after CCCP removal, treated with NPN to allow NPN to accumulate intracellularly. Then, glucose was added to initiate efflux and NPN fluorescence was measured. This assay could detect decreased efflux activity due to ΔacrB [58] (S3F Fig). The wild-type, ΔmutL, and ΔmutS strains all showed equal rates of NPN efflux following glucose addition (Fig 3D), demonstrating that increased efflux rates are not responsible for the resistance observed. To further confirm that efflux is not required, we tested whether resistance occurred in a tolC mutant, which makes at least eight tripartite efflux pumps inactive [59], and found that the increased resistance from loss of MMR genes occurred even in the absence of TolC (Figs 3E, S3G). Finally, to confirm that we were not observing heteroresistance, we performed E-tests [31] with streptomycin and gentamicin with the ΔmutL and ΔmutS strains and observed no colonies in the zone of clearance (Figs 3F, S3H). We also tested whether several stress responses (see S4 Fig for examples) were necessary for the increased resistance induced by loss of MMR genes and did not find any links to these pathways. Overall, these data eliminate the most common mechanisms conferring broad antibiotic resistance and suggest that the cells lacking MMR genes may demonstrate resistance due to a more complex physiological change.

Reported rates of homoeologous recombination correlate with increased antibiotic resistance

Since MMR can repair DNA damage beyond mismatches, we hypothesized that the resistance we observe could occur through activation of the SOS response, E. coli’s cellular stress response to severe DNA damage. The SOS response is activated when the cell senses an accumulation of single-stranded DNA [60]. The presence of RecA bound to single-stranded DNA causes then the self-cleavage of LexA (the SOS transcriptional repressor) which removes LexA from the SOS boxes in the promoters of regulated genes, derepressing transcription [61–63]. To inactivate the SOS response by preventing the release of the SOS repressor LexA from SOS boxes, we transduced a non-cleavable lexA allele [64] into our ΔbamE, ΔbamE ΔmutL, and ΔbamE ΔmutS strains. The increased resistance from ΔmutL and ΔmutS was not changed when the non-inducible lexA allele was introduced (S5A Fig), demonstrating that the SOS response does not play a role in the pathway linking loss of MMR genes to increased levels of antibiotic resistance. This is consistent with the lack of similar phenotypes caused by deletion of genes in other DNA repair pathways (S5B Fig).

In investigating mechanisms for the effect of MMR inhibition on antibiotic resistance independent of mutations or DNA damage, we considered that inhibition of MMR leads to increased rates of homoeologous recombination as MutS and MutL can block RecA-mediated strand exchange when nucleotide mismatches occur [65–67]. However, the change in rates of recombination is not equal for all MMR genes. Inhibition of MMR via ΔmutS or ΔmutL increases recombination rates by 735-fold*,* while ΔmutH and ΔuvrD increase rates 22-fold and 5-fold, respectively, with simultaneous inhibition of MutH and UvrD producing similar phenotypes to loss of MutL or MutS [65–67]. Interestingly, aside from UvrD, which also functions in nucleotide excision repair and in resolution of DNA replication-transcription conflicts and so may have secondary effects [68,69], the pattern of increased rates of homoeologous recombination is similar to the pattern of increased antibiotic we observed with the deletions of the various MMR genes (Figs 4A, S6A), with lower resistance observed for ΔmutH and, for some antibiotics, ΔuvrD, and resistance equal to ΔmutL and ΔmutS observed for the ΔmutH ΔuvrD strain. This caused us to ask whether recombination was involved in increased antibiotic resistance with loss of MMR genes [70].

Increased antibiotic resistance with loss of MMR genes is linked to homologous recombination.(A) By EOP, resistance from loss of MutL and MutS is greater than that of MutH and often UvrD; however, resistance from loss of both MutH and UvrD is similar to that of MutL and MutS. These changes correlate well with the reported rates of increase in homologous recombination rates MMR mutants. EOP was quantitated and normalized to the wild-type strain. Data are shown as the mean of at least three biological replicates ± the SEM with individual data points. (B) ΔrecA, ΔrecB, and ΔrecC, which reduce homologous recombination rates, result in decreased antibiotic resistance in both the wild type and MMR gene deletion backgrounds for most of the assayed antibiotics. EOPs were quantitated as in (A). * p < 0.05, ** p < 0.005, *** p < 0.0005 compared to wild type; ‡ p < 0.05, ‡‡ p < 0.005 compared to parent strain by Mann Whitney test.

If the increase in recombination rates in strains with MMR gene deletions is important for antibiotic resistance, then strains with reduced recombination rates should have lower resistance to antibiotics. Therefore, we combined deletions of MMR genes with deletions of recA, required for most recombination [71], as well as recB and recC, which code for proteins forming a complex with endonuclease and helicase activity that initiates recombination in the RecBCD pathway, the main homologous recombination pathway in E. coli [72]. RecB acts as a helicase and endonuclease in the RecBCD complex, while RecC is a DNA duplex splitter and binds to Chi sites, causing a conformational site allowing formation of a 3′ DNA end [72]. RecA is the DNA strand exchange protein that binds to single-stranded DNA and then brings to a homologous region of DNA to form a trimeric DNA complex [73]. ΔrecB or ΔrecC strains have an approximately 100-fold lower rate of chromosomal recombination compared to wild type, while ΔrecA decreases rates of recombination by more than 2000-fold [74–80]. We observed that while the ΔmutL or ΔmutS single mutants displayed increased resistance compared to the wild type, the ΔrecA, ΔrecB and ΔrecC single mutants were less resistant than wild type to some of the antibiotics (Figs 4B, S6B). When we coupled ΔmutL or ΔmutS with ΔrecA, ΔrecB, or ΔrecC, we observed a decrease in resistance compared to the ΔmutL or ΔmutS strains which was especially strong for the ΔrecC strains and for streptomycin. We also constructed triple mutants of ΔyhdP, ΔmutL, and the rec mutants (S6C Fig); however, genetic interactions between ΔyhdP and the rec mutants prevented these data from providing further clarity. Together, these data demonstrate that homologous and/or homoeologous recombination is important for antibiotic resistance and suggest that the increase in resistance to several antibiotics caused by loss of MMR genes is in part due to increased recombination rates.

Loss of MMR genes leads to cell lysis

Our data show that there is not a change in OM permeability in a ΔmutL or ΔmutS strain. However, we wondered whether there might be a change to cellular integrity in the MMR inactivated strains. We used chlorophenyl red-β-D-galactopyranoside (CPRG) to test for this integrity, as has been previously described [81]. When CPRG contacts β-galactosidase either in the cytoplasm or the extracellular environment, the bond between chlorophenyl red and galactose is cleaved, causing a color change from yellow to red. This assay has previously identified strains with elevated rates of lysis, such as ΔelyC [81]. We grew our MMR gene deletion strains in the presence of IPTG, to induce lac operon expression, and CPRG and assayed CPRG cleavage. The ΔelyC strain served as a positive control [81]. The MMR gene deletion strains all showed increased CPRG activity when compared to the wild-type strain (Figs 5A, S7A). Moreover, this increase in CPRG activity with ΔmutL or ΔmutS persisted even in OM deficient strain backgrounds including ΔbamE, ΔyhdP, and ΔpldA. As our previous data show that the OM is not more permeable in either the ΔmutL or ΔmutS strains (Fig 3C), these data suggest more CPRG is in contact with β-galactosidase because β-galactosidase is escaping the cell (i.e., through lysis). Therefore, we decided to investigate whether the strains with deletions of MMR genes were exhibiting some level of lysis.

Loss of MMR genes causes increased cell lysis.(A) CPRG is a poorly cell-permeable, colorimetric β-galactosidase substrate that must encounter β-galactosidase, either through increased OM permeability or cell lysis, to be cleaved releasing chlorophenol red. Deletion of MMR genes causes increased CPRG cleavage, suggesting either increased OM permeability or cell lysis. (B) To assay lysis, immunoblots detecting the cytoplasmic protein GroEL were performed on the TCA precipitated supernatant of overnight cultures from an equal culture OD600. Strains with deletions of MMR genes show higher levels of GroEL in the supernatant than wild-type cells do, demonstrating increased lysis. However, this lysis was not decreased by loss of RecB or RecC. Quantification of GroEL levels is shown as mean of three biological replicates ± the SEM. (C) Necrosignaling may play a role in resistance to some assayed antibiotics. AcrA released from lysing cells is sensed by neighboring cells to induce a decrease in antibiotic resistance [82,83]. The change in resistance from loss of necrosignaling (ΔacrA) was compared to that of loss of the AcrAB-TolC efflux pump (ΔacrB). While ΔacrA did not change levels of vancomycin or gentamicin resistance in the ΔmutL and ΔmutS backgrounds, ΔacrA resulted in significant loss of resistance to streptomycin. EOP was quantitated and normalized to the wild-type strain. Data are shown as the mean of at least three biological replicates ± the SEM with individual data points. * p < 0.05, ** p < 0.005 compared to wild type; ‡ p < 0.05 compared to parent strain by Mann Whitney test.

We performed western blots on the precipitated supernatant from overnight cultures to detect the presence of the cytoplasmic chaperonin protein GroEL in the culture media. We observed higher GroEL levels in the culture supernatant the MMR gene deletions when compared to the wild type (Fig 5B). Quantitating the GroEL levels showed they were 6- to 14-fold higher in the MMR mutants than in wild type. We wondered whether the lysis was due to illegitimate homologous recombination and so tested lysis in strains with ΔmutL or ΔmutS combined with ΔrecB and ΔrecC. However, the ΔrecB and ΔrecC strains showed higher levels of lysis than ΔmutL or ΔmutS caused, complicating analysis (Fig 5B). Nevertheless, ΔmutL or ΔmutS did not cause any additional increase in lysis in the ΔrecB and ΔrecC strains, suggesting lysis in the MMR gene deletion strains may be linked to increased recombination. As the double mutant strains are more sensitive to some antibiotics than strains lacking MMR genes alone, these data also suggest that, if lysis is involved in resistance to these antibiotic, lysis is not sufficient, or a specific level of lysis is necessary to promote resistance.

We further wondered whether the antibiotic resistance phenotypes we observed might be due to the phenomenon of “necrosignaling” recently described in swarming E. coli [82,83]. This pathway relies on AcrA, the lipoprotein component of the AcrAB-TolC RND family efflux pump [84,85]. Dead cells release a necrosignal in the form of AcrA which interacts with the OM of the population of the swarm that has yet to encounter the antibiotic. The resistance of the signaled cells to the antibiotic is increased due to the upregulation of efflux pumps and metabolic changes to the cells. We constructed ΔmutL or ΔmutS coupled with ΔacrA and ΔacrB and investigated whether these mutations would affect antibiotic resistance. Changes in resistance based on efflux should be equal between the ΔacrA and ΔacrB mutants, while necrosignaling would be affected only by ΔacrA. Resistance to vancomycin, streptomycin, and gentamicin were not strongly affected by the loss of AcrB (Figs 5C, S7B). ΔacrA did not change vancomycin or gentamicin resistance in the MMR deficient backgrounds; however, streptomycin resistance was significantly reduced in the MMR ΔacrA double mutants (Figs 5C, S7B). Therefore, both necrosignaling and rates of recombination may play a role in increasing resistance to some antibiotics in MMR mutants. These data demonstrate that loss of MMR causes some level of cell lysis within the cell population and suggest this lysis may affect resistance to some antibiotics through a necrosignaling pathway.

Discussion

Here, we present evidence that a novel pathway acts to increase population-wide antibiotic resistance in E. coli and S. Typhimurium when MMR is inhibited through deletion of MMR genes*.* This pathway acts on antibiotics of many classes, sizes, and mechanisms of action; moreover, it is independent of increased mutation rates. This increase in resistance can be observed in wild-type strains as well as in various mutants with increased OM permeability. Increased mutations rates in hypermutator strains such as MMR mutants have been shown to be associated with antibiotic resistance [30]. The pathway we have described provides a mechanism by which MMR-inhibited hypermutator strains could survive in the presence of antibiotics, giving them time to develop mutations causing true resistance. This also suggests that the proportion of MMR mutants in a population may start to expand under antibiotic treatment due to their intrinsic resistance, providing increased opportunity for antibiotic specific mutations conferring high-level resistance to occur. Previous work has shown mutations causing low-level resistance can lead to a greater chance of development of further resistance mutations, allowing strains to cross the clinical breakpoint into being resistant [7,9–12]. Therefore, this pathway mediated by loss of MMR genes, and the low-level resistance it causes, could be an important contributing factor to the development and spread of resistance alleles.

The pathway we have described causes an increase in resistance to a broad range of antibiotics via loss of MMR genes, independent of the strains’ increased mutation rates. We can confidently say the resistance is independent of mutation rates because we were able to transduce MMR gene deletions into multiple strains and backgrounds and observe a similar increase in resistance. Furthermore, we can differentiate between resistance mutations occurring within a population and a strain with increased resistance through Kirby-Bauer assays where population-level resistance determines the zone of clearance around the disc, while spontaneous mutants or heteroresistance within the population appear as colonies within the zone of clearance. Furthermore, we were able to restore mutS to our deletion strains both before and after selection and restore resistance to wild-type levels. Finally, loss of the Dam methylase has the same effect on mutation rates as loss of MMR genes but does not confer increased antibiotic resistance, demonstrating increased mutation rates are not sufficient for increased resistance. In addition to this robust evidence, the increased mutation rates should not be sufficient to cause a consistent increase in resistance between many independent strains. For example, the rate of spontaneous base-pair substitutions of E. coli K-12 is 2 × 10^-10^ mutations per nucleotide per generation [86,87]. When MMR is inactivated, there will be a 15–20-fold increase in mutations due to the activity of repair pathways mechanisms of repair that cannot differentiate between daughter and parental strands of DNA [88], corresponding to approximately 1 mutation per 75 generations of growth [87,88]. Given our attempts to avoid extended culture time (i.e., minimizing time in stationary phase) and passaging of these strains (i.e., by streaking fresh plates from glycerol stocks immediately before each experiment), this mutation rate would be insufficient for resistance mutations to consistently occur and spread throughout the cell population before resistance is assayed.

The independence of the pathway from increased mutation rates opens new questions about the relationship between DNA repair, mutation rates, and antibiotic resistance, specifically the mechanism linking loss of MMR genes to increased antibiotic resistance. Our data provide some insights into this mechanism, although elucidating the full mechanism is likely to take many years of work. Our data correlates the increased resistance to some antibiotics to increased recombination rates and suggest that some level of cell lysis within the population may be involved in resistance as well. Without MMR genes, cells will have higher rates of homoeologous recombination, where sequences between recombining strands differ. Thus, there will be a higher probability of illegitimate or unsuccessful recombination. We believe that this may lead to low levels of lysis within the population that, in turn, increase antibiotic resistance. This increased resistance could be through necrosignaling causing physiological changes in surviving cells [82,83], which, as we have demonstrated, contributes to changes in streptomycin resistance. For other antibiotics where necrosignaling does not appear to play a role, antibiotics could be exposed to and bind their targets outside of cells, lowering their effective concentration. In this way, interactions between cell debris and antibiotics could lower the amount of the antibiotic that can enter living cells. One example of this type of extracellular interaction of an antibiotic with its target leading to resistance comes from a spontaneous vancomycin resistance mutation in E. coli that causes it to attach peptidoglycan subunits to its LPS [89]. Vancomycin interacts with the cell surface instead of penetrating the OM, allowing the mutant to grow on much higher concentrations of vancomycin than the wild type.

It is also possible that illegitimate or unsuccessful recombination increases antibiotic resistance independently from the increased lysis we observe in strains with deletions of MMR genes. For instance, these events may initiate a signaling cascade that causes physiological changes, increasing the cells’ antibiotic resistance. Transcriptional changes leading to multidrug resistance have been associated with bacterial stress responses including those mediated by MarA, SoxS, and SdiA [90,91]. We previously conducted RNA sequencing of untreated MMR mutants but did not resolve any transcriptional changes that were required for changes to antibiotic resistance. However, it may be that, given the sporadic nature of recombination events, single cell RNA-Seq will be required for investigating this possibility. In addition, it may be that the illegitimate recombination events inducing transcriptional changes in MMR mutants occur specifically after antibiotic treatment. We are currently investigating this possibility. Any mediating steps we uncover in this pathway would be potential targets for decreasing the development of resistance mutations.

The RecBCD pathway and UvrD can contribute to the restart of DNA replication after a replication-transcription conflict through recombination repair or through replication fork reversal [69,92]. In addition, RecBCD cleaves DNA ends removing large segments of DNA in the alternative-end joining pathway that allows repair of double-strand DNA breaks without homologous recombination and plays a role in defense from phage through degradation of linear DNA not containing Chi sites [93,94]. Given the comparatively strong effect of ΔrecC on the phenotypes of the MMR mutants despite its weaker effect on homologous recombination than ΔrecA, one of these pathways could be involved in the antibiotic resistance we observe. However, the stronger effect of ΔrecC than ΔrecA may be because recombination rates are brought closer to wild type with the combination of ΔrecC and the MMR mutants than in the ΔrecA strain where recombination is largely absent. Finally, it is possible that antibiotic treatment induces recombination events that directly increase resistance (e.g., through gene duplication events); however, this mechanism would not repeatedly produce equal levels of antibiotic resistance and so we consider this a less likely possibility.

Much like MMR, the links proposed between recombination and antibiotic resistance have mainly been related to the spread of resistance alleles through pathways that require recombination (e.g., conjugation or transduction) [71]. More recently, DNA repair via homologous recombination has been linked to persistence after fluoroquinolone treatment [95]. Our data show that the effects of recombination rates on antibiotic resistance can extend beyond antibiotics that target DNA replication or directly cause double-strand DNA breaks. This suggests that recombination may be playing a larger role in antibiotic responses—perhaps through repair of DNA damage that occurs due to metabolic changes during antibiotic treatment [96,97]. It has been well studied that accumulation of DNA damage can lead to stress response activity [61,98–100]. It may be that DNA damage occurring during antibiotic treatment works in conjunction with increased homoeologous recombination to activate further resistance. These are intriguing areas for further investigation.

We have shown a promising link between decreased rates of recombination and decreased resistance to antibiotics, and a similar mechanism may extend beyond E. coli. Although our mechanistic studies have been performed on E. coli, there is some evidence that loss of MMR genes leads to increased rates of recombination in S. Typhimurium [66,67]. Given that the resistance we observe does not change OM permeability, it is tempting to speculate that this resistance might extend to gram-positive species. MMR is conserved within eukaryotic and prokaryotic systems [101,102]. More importantly, the pattern of suppression of homoeologous recombination via MMR is also conserved through prokaryotic and eukaryotic systems, meaning loss of MMR genes, regardless of system, could result in stress responses and adaptation to stress. For instance, studies in yeast have revealed that suppression of homoeologous recombination occurs via a mechanism that involves RecQ family helicase SGS1 and MutSα [103,104]. Elucidating this pathway further will not only provide a more detailed perspective on how increased antibiotic resistance arises, but could also inform how other non-antimicrobial resistance (temperature, pH, oxidative, and osmotic stress) occurs with MMR mutants in different organisms.

Materials and methods

Bacterial strains and growth conditions

All strains used in this study are listed in S3 Table. Cultures were grown in LB Lennox media at 37°C. Knockout strains in E. coli K-12 MG1655 were constructed with Keio collection alleles [105] using P1vir transduction [106]. The pCP20 plasmid, which encodes a FLP recombinase-FRT system to remove antibiotic resistance cassettes, was used to remove resistance cassettes after transduction [35]. New alleles in S. Typhimurium and E. coli were constructed with λ-Red recombineering using the pKD46 plasmid using the indicated primers (S4 Table) and the kanR cassette from pKD13 or the cmR cassette from pKD3 as a template, as has been described [35]. In S. Typhimurium, the recombineering plasmid was cleared by growth at 37°C. For E. coli strains, newly constructed alleles were moved to clean backgrounds using P1vir transduction.

Complementation plasmids for mutL and mutS were constructed by using the mutL_fwd/rev and mutS_fwd/rev primers (S4 Table) to amplify mutL or mutS, respectively, from genomic DNA. pBAD33 [107] was amplified using the pBAD-fwd and pBAD-rev primers and the fragments were combined with HiFi Assembly (New England Biolabs) as per the manufacturer’s instructions. Before experiments, strains were streaked for single colonies from glycerol stocks, and individual colonies cultured immediately for experiments. MMR mutant strains were not stored after growth or passaged. When necessary for plasmid maintenance, cultures were supplemented with 25 μg/mL kanamycin (Gold Biotechnology) or 20 μg/mL chloramphenicol (Gold Biotechnology). For assay of resistance phenotypes, LB was supplemented with the indicated concentrations of vancomycin (Gold Biotechnology), streptomycin (Gold Biotechnology), gentamicin (Gold Biotechnology), rifampicin (Gold Biotechnology), bacitracin (Gold Biotechnology), novobiocin (Gold Biotechnology), or kanamycin (Gold Biotechnology).

Minimum inhibitory concentration assay

Strains were grown overnight in LB. Then, cultures were normalized to an OD of 0.1 and further diluted 1:1000 in LB. Next, cultures were pipetted into a 96-well plate at a volume of 98uL. Using a multichannel pipette, 2uL of 1.5-fold antibiotic dilutions were added to the cultures. Strains were incubated at 37°C overnight. The next day, the OD_600_ was read using a BioTek Synergy H1 plate reader. The lowest concentration where growth is inhibited was considered to be the MIC. The geometric mean of the MICs from three independent experiments was calculated.

Efficiency of plating assays (EOP)

Cultures were grown overnight in LB. 200 µL of overnight culture were pipetted into 96-well plates then serial diluted by a factor of 10. Dilutions were then plated onto the appropriate plates using a 48 well pin tool and the plates incubated overnight at 37°C. For quantification, the log10 efficiency of plating was calculated as normalized to the growth of the wild-type strain on the indicated antibiotic. Wherever possible, quantification was confirmed by comparing several antibiotic concentrations. These concentrations centered around 125 mg/L for vancomycin, 2 mg/L for gentamicin, 6 mg/L for streptomycin, 7.5 mg/L for rifampicin, 1000 mg/L for bacitracin, and 5 mg/L for kanamycin.

Kirby-Bauer and E tests

Strains were grown overnight. After overnight growth, 100 µL of culture was diluted into 3 mL of melted LB top agar and then the mixture was poured onto a LB agar plate until the top agar fully covered the plate. Once solidified, antibiotic discs (BD BBL Sensi-Disc) were placed on the plate and the plate was incubated overnight at 37°C. For antibiotics that require higher amounts than provided by commercial discs, antibiotics in the following amounts were added to blank discs and dried before placing on the plate: vancomycin 300 µg, erythromycin 150 µg, novobiocin 200 µg, bacitracin 1 mg, or rifampicin 20 µg. The following day, zones of inhibition were measured in 3 different directions and averaged. For E tests, plates were prepared with culture and top agar as described above and then MIC test strips (Liofilchem) were placed on the solidified top agar. Plates were incubated overnight at 37°C and then imaged.

Outer membrane permeability and efflux assays

The nitrocefin (Apexbio Technology) cleavage assay was performed on the indicated strains carrying a bla encoding plasmid as has been previously described [55]. Efflux activity was assay based on efflux of NPN (Thermo Scientific) from cells which had previously been treated with CCCP (carbonyl cyanide m-chlorophenyl hydrazone, Thermo Scientific) following addition of glucose as has previously been described [55].

CPRG assay

CPRG assays were performed as previously described with minor variations [27,81]. LB agar plates were supplemented with 20 g/L of CPRG and 50 µM IPTG. For plate assays, bacteria were streaked out onto a CPRG LB agar plate and let to grow overnight in the dark at room temperature. For quantification, liquid CPRG assays were performed by growing strains overnight at 37°C with 50 µM IPTG. 200 µL of cells were spun down at 3700 rpm for 2 minutes in a 96-well plate. 100 µL of supernatant were transferred into a new 96-well plate, 100 µL of CPRG were then added and plate was incubated at 37°C for 1 hour. Chlorophenol red absorption was read at a 575 nm wavelength.

Immunoblot analysis for lysis via GroEL detection

The trichloroacetic acid (TCA) precipitation protocol was adapted from Ruiz, et al. [28]. Cultures were grown overnight in LB at 37°C. After growth, 1.2 mL of culture at a normalized OD_600_ was spun down at room temperature, 16,000 x g for 2 minutes. The supernatant was filtered through a 0.22 µm syringe filter, then 900 µL of filtered supernatant was transferred to a new microcentrifuge tube and proteins were precipitated with 100 µL of 100% TCA. TCA-precipitated proteins were then resuspended in 50 µl 1XSDS loading buffer. 20 µL of sample was loaded onto a 12% SDS-PAGE gel and the gel was run for 1 hour at 115 volts. Proteins from the gel were transferred to a nitrocellulose membrane. The GroEL chaperonin was detected with a GroEL primary antibody (Sigma) at a 1:30,000 dilution. The secondary antibody was goat anti-rabbit conjugated to horseradish peroxidase (Prometheus) at a 1:100,000 dilution. Detection of the signal was completed by using Prosignal Pico ECL (Prometheus) and Prosignal enhanced chemiluminescence (ECL) blotting film (Prometheus). Quantification of proteins levels was performed with ImageJ.

Raw data for Figs 1C, 3C, 3D, S2A, and S6A and S1 Table are provided in S1 Dataset.

Supporting information

S1 FigDNA Mismatch Repair Pathway.A nucleotide mismatch formed during DNA replication is first recognized by MutS. MutS then recruits MutL and forms a heterodimer complex at the site of the mismatch. This complex recruits MutH which scans to the nearest GATC site and makes a nick on the unmethylated strand. UvrD unwinds the DNA allowing the exonuclease to enter and degrade the DNA strand past the point of the mismatch with single stranded binding protein (SSB) stabilizing the DNA. DNA polymerase resynthesizes the DNA strand and DNA ligase ligates the remaining nick. Finally, DAM methylates the nascent DNA strand.(PNG)

S2 FigLoss of MMR does not cause a growth defect.(A) Strains with deletions in MMR genes and wild-type cells were grown at 37°C for 12 hours. No growth differences between the strains were observed. Data are representative of 3 independent experiments. Mean ±Std Dev is shown. (B) Plasmid-based complementation of MMR gene deletions returned antibiotic resistance to wild type levels, as assayed by EOP. mutL or mutS were expressed from the arabinose inducible P_BAD_ promoter in the indicated strain backgrounds. EOP was normalized to wild-type empty vector and shown as mean ± the SEM with individual data points from three biological replicates. * p<0.05 by Mann Whitney test. (C) Representative images related to Fig 1E. A Tn10 linked to mutS was transduced into two ΔmutS strains and resulting mutS+ and ΔmutS transductants were tested for their antibiotic resistance. (D) An efficiency of plating assay on gentamicin was performed with DB35. Then, three isolated colonies were picked and transduced with a Tn10 linked to mutS resulting in mutS+ and ΔmutS transductants. These transductants were tested for their gentamicin resistance by EOP. mutS+ transductants showed resistance equal to wild type, demonstrating resistance mutations are not developed during selection.(PNG)

S3 FigMMR mutants produce resistance in OM mutants.(A) Representative images related to Fig 3A. (B) EOPs showing MMR mutants retain antibiotic resistance in mutants of bamB, which codes for a non-essential BAM lipoprotein. (C) MMR mutants as retain resistance with a promoter down mutation of bamA, an essential member of the Bam complex. (D) Representative images related to Fig 3B. (E) Controls related to Fig 3C showing strains that express an open porin (pompCΔW103-F110) (56) and decrease porin expression (ΔompR) (57), respectively. Both of these strains show increased nitrocefin cleavage suggesting that nitrocefin reports on permeability based on porin availability and based on the lipid component of the membrane which is disrupted by the greatly decreased porin expression in a ΔompR strain (F) Controls related to Fig 3D showing strains that decrease (ΔacrB) efflux capacity decreases NPN efflux (58). (G) Representative images related to Fig 3E. (H) Representative images from face of plate showing MIC test strips.(PNG)

S4 FigIncreased antibiotic resistance with loss of MMR is independent of several stress responses.Deletions in various known stress response pathways were made to see whether they played a role in increased resistance in ΔmutL or ΔmutS strains. If these stress responses were involved in the increased resistance we observe with loss of MMR, we would expect decreased resistance when the stress response is inhibited (rcsF::kan, soxR::kan) or increase resistance with its activation (cpxP::kan). No such changes were observed.(PNG)

S5 FigIncreased antibiotic resistance does not occur with loss of the SOS response or inhibition of other DNA repair pathways.(A) The SOS response is E. coli’s response to DNA damage that leads to exposure of single-stranded DNA. Inactivation of the SOS response via lexA3 (a non-cleavable mutant of the SOS transcriptional regulator) does not change antibiotic resistance mediated by loss of MMR genes as assayed by EOP. (B) To investigate whether the increased antibiotic resistance we observe with loss of MMR occurs with other DNA repair pathways, we made deletions of several DNA repair associated genes that span multiple DNA repair mechanisms (nucleotide excision, base excision, aklyation, etc.) and found inhibition of other DNA repair pathways does not lead to a similar pattern of increased resistance to that observed with loss of MMR.(PNG)

S6 FigIncreased resistance from loss of MMR correlates with reported changes in homologous recombination rates.(A) Images related to Fig 4A. (B) Images related to Fig 4B. (C) EOP for recombination mutants in combination with ΔmutL in a ΔyhdP background was assayed. These mutations showed effects on resistance both with and without ΔmutL. EOPs were normalized to wild type and data are shown as the mean of at least three biological replicates ± the SEM with individual data points. * p<0.05, *** p<0.0005 compared to wild type; ‡ p<0.05, compared to parent strain by Mann Whitney test.(PNG)

S7 FigLoss of MMR causes increased lysis.(A) The indicated strains were grown overnight with IPTG and then culture supernatants were harvested and incubated with CPRG before chlorophenol red absorbance was assayed to quantitate CPRG activity. An ∆elyC strain serves as a positive control. The ΔmutL strain showed increased CPRG activity compared to the wild type. Data is representative of at least 3 independent experiments. Mean ± the Standard Deviation is shown. (B) Representative images related to Fig 5C.(PNG)

S1 TableLoss of MMR increases antibiotic minimum inhibitory concentration (MIC).(PDF)

S2 TableStreptomycin resistance mutations are not observed in cells without mutL.(PDF)

S3 TableStrains used in this study.(PDF)

S4 TablePrimers used in this study.(PDF)

S1 DatasetRaw data for absorbance assays, fluorescence assays, and antibiotic resistance assays.Datasets are labeled with the Fig to which they correspond.(XLSX)